|Publication number||US8103051 B2|
|Application number||US 12/273,479|
|Publication date||Jan 24, 2012|
|Filing date||Nov 18, 2008|
|Priority date||Jan 11, 1999|
|Also published as||US6442283, US6751337, US7454034, US20030095685, US20050025334, US20090304226|
|Publication number||12273479, 273479, US 8103051 B2, US 8103051B2, US-B2-8103051, US8103051 B2, US8103051B2|
|Inventors||Ahmed Tewfik, Mitchell D. Swanson, Bin Zhu|
|Original Assignee||Digimarc Corporation|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (60), Non-Patent Citations (26), Referenced by (2), Classifications (12), Legal Events (6)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This application is a continuation of U.S. application Ser. No. 10/869,178, filed Jun. 15, 2004 (now U.S. Pat. No. 7,454,034), which is a continuation of U.S. application Ser. No. 10/229,382, filed Aug. 26, 2002 (now U.S. Pat. No. 6,751,337), which is a continuation of U.S. application Ser. No. 09/228,224, filed Jan. 11, 1999 (now U.S. Pat. No. 6,442,283). These patents are incorporated herein by reference.
This invention relates generally to multimedia data, and more particularly to multimedia data embedding.
With the increasingly popularity of multimedia-capable computers, and the digitalization of multimedia in general, the importance of multimedia data embedding has become more important. In one type of multimedia data embedding, a key, also know as a watermark, is embedded into multimedia data, a process which is known as watermarking. This allows questions of ownership of a given piece of multimedia data—which may be widely distributed by virtue of the Internet, for example—to be resolved, by attempting to decode the key from the multimedia data. That is, by watermarking multimedia data, the data owner can determine whether a suspect piece of multimedia data is his or hers by determining whether the watermark is present in the suspect data.
For example, a record company, prior to making its music selections available on the Internet for widespread purchase and use, can first watermark the data representing a music selection. If a site on the Internet is providing bootleg copies of the music selections, but claims that the copies are not in fact owned by the record company, the company can prove that they are indeed owned by it by showing that the watermark is present in the bootleg copies. Therefore, watermarking has applicability to audio multimedia, as well as other types of multimedia, such as image and video multimedia.
The invention provides methods for embedding and decoding data embedded in media signals and related software implementations.
One aspect of the invention is a method for embedding data into a media signal. The method receives a media signal, divides the media signal into blocks of samples, and calculates a function of the samples in the blocks, including transformations of samples in the blocks to corresponding block values. A processor uses the block value to determine a factor for samples in the blocks to be multiplied by the samples so that when a data embedding function is evaluated for the block, an output of the data embedding function corresponds to a data value representing desired digital data embedded in the block.
Another aspect of the invention is a method of decoding data embedded in a media signal. Like the embedder, the decoder divides the media signal into blocks of samples and calculates a function of the samples in the blocks, including transformations of samples in the blocks to corresponding block values. A processor processes the block value to evaluate a data embedding function to determine digital data embedded in the block.
In one digital watermark embodiment, a decoder projects a digitally watermarked signal into a direction according to a key. It applies a weighting function to the projected signal to compute a projected signal in which parts of the digitally watermarked signal that are more robust to distortion are weighted more than parts that are less robust to the distortion. The method recovers embedded auxiliary data symbols from the projected signal by quantizing the projected signal to determine a binary symbol associated with a quantization of the projected signal.
In the following detailed description of exemplary embodiments of the invention, reference is made to the accompanying drawings which form a part hereof, and in which is shown by way of illustration specific exemplary embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention, and it is to be understood that other embodiments may be utilized and that logical, mechanical, electrical and other changes may be made without departing from the spirit or scope of the present invention. The following detailed description is, therefore, not to be taken in a limiting sense, and the scope of the present invention is defined only by the appended claims.
Some portions of the detailed descriptions which follow are presented in terms of algorithms and symbolic representations of operations on data bits within a computer memory. These algorithmic descriptions and representations are the means used by those skilled in the data processing arts to most effectively convey the substance of their work to others skilled in the art. An algorithm is here, and generally, conceived to be a self-consistent sequence of steps leading to a desired result. The steps are those requiring physical manipulations of physical quantities. Usually, though not necessarily, these quantities take the form of electrical or magnetic signals capable of being stored, transferred, combined, compared, and otherwise manipulated. It has proven convenient at times, principally for reasons of common usage, to refer to these signals as bits, values, elements, symbols, characters, terms, numbers, or the like. It should be borne in mind, however, that all of these and similar terms are to be associated with the appropriate physical quantities and are merely convenient labels applied to these quantities. Unless specifically stated otherwise as apparent from the following discussions, it is appreciated that throughout the present invention, discussions utilizing terms such as “processing” or “computing” or “calculating” or “determining” or “displaying” or the like, refer to the action and processes of a computer system, or similar electronic computing device, that manipulates and transforms data represented as physical (electronic) quantities within the computer system's registers and memories into other data similarly represented as physical quantities within the computer system memories or registers or other such information storage, transmission or display devices.
Referring first to
In block 100, a vector x is received that represents multimedia data, such as audio, image, or video data; the invention is not so limited. In block 100,
x=[x(0)x(1) . . . x(N−1)]
and denotes a vector of N data samples.
In block 102, a vector p is received that represents a pseudo-random sequence. The vector p is the key or watermark that is to be embedded in the vector x. More specifically,
p=[p(0)p(1) . . . p(N−1)]
and represents a cryptographically secure pseudo-random sequence generated from a one-way function and a key, as known within the art.
In block 104, a vector x′ is generated, in which the vector p is embedded into the vector x. The vector x′ is the watermarked data, or the data into which the key has been embedded. More specifically, the new data vector
x′=[x′(0)x′(1) . . . x′(N−1)]
is generated by adding a second vector to the data vector x producing the new data vector
where a is a perception-based scaling factor and vector q is a perceptually weighted pseudo-random sequence. Both components a and q are perception-based to insure that x and x′ are indistinguishable to the human audio or visual systems for audio and image/video data, respectively. The computation of q and a depend on the pseudo-random sequence p and a weighting mechanism as described below. Note that the new data vector in (1) may be represented by
where N orthogonal pseudo-random sequences qn are employed. Each term anqn is used to carry one bit of information.
Referring next to
In block 200, a multimedia vector x′ is first received, from in which a key p has been embedded into a multimedia vector x. In block 202, the key p is decoded from the vector x′. To decode the embedded data, a scaled inner product between the new data vector x′ and the pseudo-random sequence p is computed
where w is a vector of weights as described below. The value of T dictates the quantization step of the algorithm and is dependent on the weighting-mechanism employed. The first term on the right hand side of (2) is referred to as the residual. It represents the projection of the original data sequence x onto the pseudo-random direction p weighted by w. The second term on the right hand side of (2) is the projection of the shaped pseudo-random sequence q with the pseudo-random direction p weighted by w.
As described in (1), the second term carries the embedded information. The residual R
is known. Using this knowledge, a variable d is defined where
where b is the data bit to embed. The variable a in (1) is computed as
Substituting (3) and (4) into (2), the embedded data can be recovered without error in an environment without distortion.
In at least some embodiments of the invention, factors to consider when embedding data in audio, the weighting function w, and the shaped pseudo-random sequence q are described below.
Audio Data Hiding Considerations
The audio data hiding algorithm works by making perceptually insignificant modifications to the audio samples. The audio signal is modified in blocks of size Nb, i.e., Nb consecutive samples of the audio are processed at the same time. In one implementation, the blocks are non-overlapping. However, overlapping blocks may be used.
The data embedding algorithm described above is computed in the discrete cosine transform (DCT) domain for audio signals. Due to the presence of efficient Fast Fourier Transform methodologies known in the art, Nb is typically selected as a power of 2, e.g., Nb=1024. The size of the block is controlled by several factors. Since audio characteristics may change rapidly, the block size should be small to keep modifications localized in time. Smaller block sizes are also preferred during the decoding process during synchronization. However, the block size should be large enough to provide a high frequency resolution. The DCT frequency resolution is computed as
where fs is the sampling frequency of the audio signal. For Nb=1024 and fs=44100 Hz, the frequency resolution is fd=21.53 Hz.
A length Nb block of audio samples produces Nb DCT coefficients. In the standard implementation, the DCT spectrum is generally segmented into smaller subbands. In particular, a subset of length N<Nb of the DCT coefficients is used to embed each data bit as described in (1). Each subband may consist of a different number of DCT coefficients. For example, the spectrum may be segmented into three bands in the frequency ranges of 1000-4000 Hz, 4000-8000 Hz, and 8000-15000 Hz.
The audio data embedding procedure also includes the MPEG psychoacoustic masking model 1 or 2, as known in the art, for checking tolerable error levels. The MPEG masking model is used to verify the perceptual quality of (1). Components of the embedded data signal may be scaled or clipped to meet the requirements of the masking model.
The Pseudo-Random Sequence
In one embodiment, one or two random keys x1 and x2 (i.e., seeds) are used from which a pseudo-random sequence p can be generated, by using a suitable pseudo-random sequence generator, such as described in R. Rivest, “Cryptography,” pp. 717-755, in J. van Leeuwen (ed.), Handbook of Theoretical Computer Science, Vol. 1, Ch. 13, MIT Press, Cambridge, Mass., 1990, which is hereby incorporated by reference. Only the first key, x1, is used for most data embedding applications. The second key, x2, is required for watermarking audio. It is used to make counterfeiting very difficult. Popular generators include RSA, Rabin, Blum/Micali, and Blum/Blum/Shub, as known in the art, and as described in S. Goldwasser, M. Bellare, “Lecture notes on cryptography”, preprint, July 1996: http://www-cse.ucsd.edu/users/mihir/papers/crypto-papers.html. With the proper keys, the embedded data may be extracted. Without the key(s), the data hidden in the signal is statistically undetectable and impossible to recover. Note that classical maximal length pseudo noise sequence (i.e., m-sequence) generated by linear feedback shift registers are not used to generate a pseudo-random sequence. Sequences generated by shift registers are cryptographically insecure: one can solve for the feedback pattern (i.e., the keys) given a small number of output bits p.
The noise-like sequence p can be used to derive the actual watermark hidden into the audio signal or control the operation of the watermarking algorithm, e.g., determine the location of samples that may be modified. The key x1 is author dependent. A second key, x2, is signal dependent. The key x1 is the key assigned to (or chosen by) the author. Key x2 is computed from the audio signal when the author wishes to watermark the audio signal. It is computed from the signal using a one-way hash function. For example, the tolerable error levels supplied by masking models are hashed to a key x2. Any one of a number of well-known secure one way hash functions may be used to compute x2, including RSA, MD4, and SHA, as known in the art. MD4 is specifically described in R. Rivest, “The MD4 message digest algorithm”, pp. 303-311 in Advances in Cryptology, CRYPTO 92, Springer, Tokyo, 1991, which is hereby incorporated by reference; SHA is specifically described in National Institute of Standards and Technology (NIST), Secure Hash Standard, NIST FIPS Pub. 180-1, April 1995, which is also hereby incorporated by reference. For example, the Blum/Blum/Shub pseudo-random generator uses the one way function y=g_n(x)=x^2 mod n where n=pq for primes p and q so that p=q=3 mod 4. In at least some embodiments, generating x or y from partial knowledge of y is computationally infeasible for the Blum/Blum/Shub generator.
A QR orthogonal-triangular decomposition operation is performed on the pseudo-random sequences before they are employed by the data embedding algorithm. A typical pseudo-random sequence generator creates a sequence of samples with values ranging from −1 to +1. The relative magnitudes of samples in the sequence may be on the order of 10^6, leading to spiking and poor weighting characteristics. A QR decomposition is employed to maintain a relative magnitude in the samples on the order of 0.9 to 1.1.
The Weighting Function
A number of functions to weight the pseudo-random sequence p for robustness and perceptual quality can be been employed in accordance with different embodiments of the invention. The weighting coefficients w are generally computed as a function of the data coefficients x.
One method to generate the weighting values includes computing the average of the absolute value of the data coefficients about a length Nf interval
The value of Nf is kept small to keep the averaging localized, e.g., Nf=13. Out-of-band DCT coefficients were used at the boundaries of the averaging interval.
In other embodiments of the invention, the DCT subband is segmented into critical bands as described by the MPEG psychoacoustic model 1, known within the art. Each subband consists of Nc critical bands. The varying length critical bands increase in size with frequency. Several techniques to compute the weight wi for each critical band were employed. Note that each wi is a vector of the same length as each critical band.
In some embodiments, the weighting function is computed independently for each critical band. These include the one-norm,
Each of the weights is constant over its corresponding critical band.
The purpose of the weighting function is to approximate the relative tolerable error level for each DCT component and the corresponding accuracy at the receiver. For example, a weight for a tonal critical band may be large. However, the weight relative to the tonal is small. This is designed to emulate coding algorithms which generally introduce a smaller relative error in tonal components than in non-tonal components.
As the weight estimate is required at the receiver, the weighting function is required to be robust to many distortions. The aforementioned weighting functions are designed to perform well in terms of relative error before and after distortions to the host audio signal. In most cases the value of the weighting function wi depends on several data samples.
The Shaped Pseudo-Random Sequence
The second term in the data embedding methodology (1) includes two components: a and q. The a term was defined in (4). The shaped pseudo-random sequence, q, may be computed in a variety of manners from the secure pseudo-random sequence p.
In one embodiment, the shaped pseudo-random sequence is defined as
where w consists of a weighting function defined in the previous section and * represents a component-by-component multiplication. As a result, the modification to the original data sequence is a scaled version of the pseudo-random sequence shaped by the weights.
A second approach employs a finer resolution in the modification by defining the shaped pseudo-random sequence as
In this case, the pseudo-random sequence is shaped by the absolute value of the data it is modifying. The finer resolution in this case is due to the fact that the weights wi in the previous section are constant over multiple samples, i.e., the critical band. The modification in is performed at an individual sample level.
The previously described shaping techniques for the pseudo-random sequence only take into account frequency shaping, in at least some embodiments of the invention. To insure that the data embedding algorithm avoids pre- and post-echo distortions, a temporal shaping component is introduced. Recall that the embedding methodology is computed in the frequency domain. Let xt denote the data block in the time domain. Note that the length of the time data vector is Nb, i.e., the same length as the DCT block. To account for temporal shaping, the envelope of the data signal in the time domain is generated. First the DCT of the absolute value of the data in the time domain is computed
A second DCT signal X′ is generated by retaining only the first K low frequency coefficients of X
Typically, 6<K<10 depending on the desired amount of smoothing of the envelope. An inverse DCT computation is performed on X′ generating a smoothed envelope m of the data signal.
To generate q, an intermediate signal q′ is computed as described in (5)
The time-domain representation, qt′, of the pseudo-random sequence q′ is computed by
and multiplied by the envelope m, generating a temporally-shaped pseudo-random sequence
The DCT of the temporally-shaped pseudo-random sequence is computed, resulting in the final shaped pseudo-random sequence
The value of a is then computed as described by (4). Note that the pseudo-random sequence p is shaped in both the frequency and time domains to increase the perceptual quality of the embedded data.
In one particular embodiment, a window h is introduced in (6). In particular, the window is introduced to generate a temporally-shaped and windowed pseudo-random sequence
The other calculations are not affected. A rectangular window introduces an audible blocking effect in some audio signals. A shaped window h that tapers off near the beginning and ending of the block prevents the blocking noise. Gaussian and trapeziodal windows, as known in the art, are employed by the data embedding methodology.
A different approach to compute q based on coding error is now described. In this embodiment, the coded audio signal xc at the target bit rate and coding algorithm is generated and used to obtain the coding error
The shaped pseudo-random sequence q is computed as
The approach exploits perceptual characteristics of most current audio coding algorithms. In particular, the coding error e generated by popular algorithms, e.g., Dolby's AC-3 and MPEC, as known in the art, is typically perceptually shaped.
In a related embodiment, the value of a described in (4) is modified to take into account coding error. Several different values of a are computed according to
where 0<<b<1 is employed to tweak to the value of a to the best value for the target bit rate. Each is tested by coding the embedded audio algorithm and recovering the embedded data bit. The best a is selected.
Tonal (harmonic) and non-tonal (residue) components in an audio signal have different perceptual characteristics based on the masking properties of the human auditory system. For example, the amplitude of non-tonal and low level noise components in an audio signal may be modified in the range of 50-100% without a perceptually significant change to the audio. Perceptual changes in tonal components, however, may sometimes be detected after only a 10\% change in amplitude. The strength of embedded data may be limited by maximum allowable changes to the tonal components. To enhance the audio data hiding algorithm, a tool to detect the tonal components in audio signals was developed. By separating tonal and non-tonal components, the data embedding algorithm is able to maximize the strength of each component independently.
The tonal components in an audio signal are identified using a harmonic analysis technique as described in D. J. Thomson, “Spectrum estimation and harmonic analysis,” Proceedings of the IEEE, vol. 70, no. 9, pp. 1055-1096, September, 1982, which is hereby incorporated by reference. The analysis provides an accurate estimate of the location (frequency), amplitude, and phase of harmonic components in the audio. As described in the following section, this information may be used to change the weighting function. It may also used in an alternative data hiding.
The technique described in Thomson analyzes harmonic components in an audio signal by expanding the audio signal in terms of a set of orthogonal windows called prolate sequences. The expansion is followed by a statistical F-test to determine whether a tonal component exists at a particular frequency.
The harmonic analysis is performed on segments of Nb audio samples as described in the previous section. The audio segment is multiplied (windowed) by a set of K prolate sequences. A zero-padded discrete Fourier transform of each windowed data sequence is then computed. The windowing provides K different estimates of the spectrum based on the K different prolate windows. Zero-padding is used to prevent circular wrapping and provide a high level of frequency resolution. Typically the sequence is zero-padded to 2 Nb or 4 Nb The harmonic mean at each frequency is then computed followed by an F-test statistic is then computed. The F-value at each frequency is a measure of the ratio of the estimate of the magnitude of the harmonic at that frequency to that of the non-tonal part of the spectrum. If a tonal exists at a particular frequency, the F-value will be large. A small F-value indicates that the component is non-tonal. The frequencies corresponding to tonals in the audio are obtained by finding peaks in the F-statistics. The frequency, amplitude, and phase of each tonal component are then provided to the audio data embedding algorithm. The process is repeated on each length Nb segment of the audio.
The tonal detection methodology provides the position (frequency) of each tonal component in a length Nb block of audio. With the tonal detection and audio data embedding algorithms synchronized in time and employing the same number of audio samples Nb, the data embedding algorithm uses the detected tonals occurring in each frequency band to modify the weighting function. Weights corresponding to tonal components (and frequency components near the tonal) are scaled differently than non-tonal components, let z denote the set of tonal indexes returned by the tonal detection algorithm. A modified version x′ of the data x is generated such that
where 0<B<1 is a scaling value for tonal components. Parameter q is computed as described earlier using x′ in place of x, i.e.,
As a result, the tonal and non-tonal components of the data have a different effect on the computation of q through the use of the parameter B. This allows the algorithm to modify the amplitude of tonal components, e.g., by +/−10%, differently than non-tonal components, e.g, by +/−50%. The same tonal detection algorithm is employed by the receiver to generate the appropriate weights and recover the embedded information.
An alternative embedding scheme employing tonal detection is now described. The embedding methodology first separates the original audio signal into a tonal component and a residual (non-tonal) component
x=x r +x t
This embodiment uses the frequency, amplitude, and phase information provided by the tonal detection procedure to extract the tonal components.
The data embedding methodology modifies each component in a different manner. The residual component, xr, is modified using the standard embedding methodology described by (1) Information is embedded in the tonal components, xt, by shifting the relative position (frequency) of the tonals in an audio block. The tonal may be shifted since the human ear is unable to detect a difference in frequency within 3.6 Hz for frequencies below 500 Hz, and within 0.007 f for frequencies f>500 Hz. For frequencies in Layer 1, 1000 Hz to 4000 Hz, the frequencies may change from 7 to 28 Hz, respectively. The modification is usually limited from 5 to 20 Hz to ensure perceptual quality. The modifications to the length Nb block are generally performed on a 2 Nb or 4 Nb zero-padded FFT to guarantee a high frequency resolution. The frequencies are shifted in accordance with a pseudo-random pattern.
Once the frequencies have been modified, the sinusoidal signal synthesis methodology described in R. J. McAulay, T. F. Quatieri, “Speech analysis/synthesis based on a sinusoidal representation,” IEEE Trans. On Acoustics, Speech, and Signal Processing, vol. 34, no. 4, pp. 744-754, August, 1986, which is hereby incorporated by reference, is used to reconstruct the signal from the modified tonals. The McAulay methodology tracks frequencies from block to block to avoid discontinuities in the amplitudes and phases. The reconstruction algorithm locates a tonal in the next audio block closest to the tonal in the current block. If the closest tonal in the next block is within a pre-defined frequency range, the difference is assumed to represent the varying nature of audio. If the closest tonal in the next block is out of the frequency range, the tonal in the current block is assumed to have ceased. If a tonal appears in the next block that does not occur in the current block, a new tonal is flagged for tracking.
The amplitudes of tracked tonal components from block to block are matched using linear interpolation. A cubic interpolation function is used to match phases of tracked tonal components. The tonal component of the audio signal is then reconstructed by taking the inverse Fourier transform of the modified tonal amplitudes and phases.
The residual component, modified by the original audio data embedding algorithm, is added to the tonal component. The resulting signal has data embedded in both the tonal and non-tonal components.
The receiver recovers the embedded information by separating the tonal and residual components using the tonal detection algorithm. Data embedded in the residual component is recovered using the original data detection procedure. Information embedded in the tonal components is extracted by comparing the relative positions of the tonals with the pseudo-random patterns used by the data embedding algorithm. In particular, the information bit stream recovered depends on the match of the relative positions with the appropriate set of tonal patterns.
Referring now to
The carrier signal 300 in the system is the audio or image host signal. The data embedding algorithm already described modulates the embedded data 302 (that is, the key or the watermark) with the carrier signal, as represented by block 304. The raw embed rate for mono CD quality audio at this stage in the system is 252 bits/sec (504 bits/sec stereo audio). The raw embed rate does not take into account any reduction in bit rate required for error correction or synchronization. For a 512×512 grayscale image, the raw data rate at this point is 8192 bits. The embedded bit rate may be up to 3 times higher for color images.
The next stage in the system consists of error correction and synchronization algorithms, as represented by block 306. Both are used for robustness under unknown channel conditions in one embodiment of the invention. Synchronization consists of repeatedly embedding a random pattern of bits known to the receiver. Much like the embedded data, the secure random pattern is based on encryption algorithms and may only be located by the appropriate audience. The synchronization bits reduce the raw bit rate by approximately 16%. Synchronization is essential to counteract the effects of the channel distortions that may delay, resize, rotate, crop, etc., the host signal. The receiver must be properly aligned with the embedded data for an accurate recovery of the information.
Two forms of error correction are employed by the data embedding system of
Error correction via averaging works in conjunction with the second error correction mechanism: error control coding. Error control codes use sophisticated functions to increase the reliability of digital data transmission. Error control coding works most efficiently in environments with relatively low bit error rates. Thus, error correction via averaging is an essential preprocessing step to keep the bit errors low prior to error control coding. A commonly used error control code is a block code, e.g., Hamming and BCH, as described in S. Lin, D. J. Costello, Error Control Coding: Fundamentals and Applications, Prentice-Hall Inc., Englewood Cliffs, N.J., 1991, which is hereby incorporated by reference. A block code breaks an information stream into message blocks of size k. The message block of length k is then represented by a length n codeword, where n>k. A total of n-k redundant bits are added to each message to detect and correct errors in the message introduced by the noisy channel.
Once the embedding process is done, the audio or image passes through the communication channel, as represented by block 308. The channel consists of any medium that may hold the audio or image data. The data may remain digital when transmitted through the channel, or may be converted to an analog form. For audio, this may include analog tapes, telephones, broadcast etc. For image media, the channel may include printer paper, newspapers, faxes, magazines, etc. Furthermore, any number of enhancements, coding representations, cropping, scaling, etc., may be applied to the host signal before reaching the receiver. A number of the degradation and distortions, e.g., telephone, printing, faxing, scanning, taping, can occur.
When the receiver obtains the host signal, the detection algorithm first synchronizes the received signal, as represented by block 310. Synchronization may require a search over a range of delays, scales, and rotations, to properly align the received data. Once synchronized, the embedded data is extracted and processed by the error correction mechanisms. The values obtained for each repeated bit are combined and averaged to produce a bit estimate with reduced channel error. The BCH error control code is then applied to further reduce any bit errors, as represented by block 312. This significantly decreases the chance of suffering stray bit errors. The resulting extracted data 314 thus includes bits that are properly assembled into the proper ASCII text or binary representation to reform the embedded information.
Referring finally to
Multimedia data embedding has been described. Although specific embodiments have been illustrated and described herein, it will be appreciated by those of ordinary skill in the art that any arrangement which is calculated to achieve the same purpose may be substituted for the specific embodiments shown. This application is intended to cover any adaptations or variations of the present invention. Therefore, it is manifestly intended that this invention be limited only by the following claims and equivalents thereof.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US3406344||Jul 1, 1964||Oct 15, 1968||Bell Telephone Labor Inc||Transmission of low frequency signals by modulation of voice carrier|
|US3845391||Jul 15, 1971||Oct 29, 1974||Audicom Corp||Communication including submerged identification signal|
|US4313197||Apr 9, 1980||Jan 26, 1982||Bell Telephone Laboratories, Incorporated||Spread spectrum arrangement for (de)multiplexing speech signals and nonspeech signals|
|US4425661||Sep 3, 1981||Jan 10, 1984||Applied Spectrum Technologies, Inc.||Data under voice communications system|
|US4939515||Sep 30, 1988||Jul 3, 1990||General Electric Company||Digital signal encoding and decoding apparatus|
|US4943973||Mar 31, 1989||Jul 24, 1990||At&T Company||Spread-spectrum identification signal for communications system|
|US5161210||Nov 8, 1989||Nov 3, 1992||U.S. Philips Corporation||Coder for incorporating an auxiliary information signal in a digital audio signal, decoder for recovering such signals from the combined signal, and record carrier having such combined signal recorded thereon|
|US5319735||Dec 17, 1991||Jun 7, 1994||Bolt Beranek And Newman Inc.||Embedded signalling|
|US5404377||Apr 8, 1994||Apr 4, 1995||Moses; Donald W.||Simultaneous transmission of data and audio signals by means of perceptual coding|
|US5450490||Mar 31, 1994||Sep 12, 1995||The Arbitron Company||Apparatus and methods for including codes in audio signals and decoding|
|US5613004||Jun 7, 1995||Mar 18, 1997||The Dice Company||Steganographic method and device|
|US5649054||Dec 21, 1994||Jul 15, 1997||U.S. Philips Corporation||Method and apparatus for coding digital sound by subtracting adaptive dither and inserting buried channel bits and an apparatus for decoding such encoding digital sound|
|US5774452||Mar 14, 1995||Jun 30, 1998||Aris Technologies, Inc.||Apparatus and method for encoding and decoding information in audio signals|
|US5809139||Sep 13, 1996||Sep 15, 1998||Vivo Software, Inc.||Watermarking method and apparatus for compressed digital video|
|US5822360||Sep 6, 1995||Oct 13, 1998||Solana Technology Development Corporation||Method and apparatus for transporting auxiliary data in audio signals|
|US5822432||Jan 17, 1996||Oct 13, 1998||The Dice Company||Method for human-assisted random key generation and application for digital watermark system|
|US5828325||Apr 3, 1996||Oct 27, 1998||Aris Technologies, Inc.||Apparatus and method for encoding and decoding information in analog signals|
|US5848155||Sep 4, 1996||Dec 8, 1998||Nec Research Institute, Inc.||Spread spectrum watermark for embedded signalling|
|US5889868||Jul 2, 1996||Mar 30, 1999||The Dice Company||Optimization methods for the insertion, protection, and detection of digital watermarks in digitized data|
|US5905800||Mar 25, 1998||May 18, 1999||The Dice Company||Method and system for digital watermarking|
|US5915027||Nov 5, 1996||Jun 22, 1999||Nec Research Institute||Digital watermarking|
|US5930369||Sep 10, 1997||Jul 27, 1999||Nec Research Institute, Inc.||Secure spread spectrum watermarking for multimedia data|
|US5933798||Jul 16, 1997||Aug 3, 1999||U.S. Philips Corporation||Detecting a watermark embedded in an information signal|
|US5937000||Dec 6, 1996||Aug 10, 1999||Solana Technology Development Corporation||Method and apparatus for embedding auxiliary data in a primary data signal|
|US5940135||May 19, 1997||Aug 17, 1999||Aris Technologies, Inc.||Apparatus and method for encoding and decoding information in analog signals|
|US5940429||Feb 25, 1997||Aug 17, 1999||Solana Technology Development Corporation||Cross-term compensation power adjustment of embedded auxiliary data in a primary data signal|
|US5945932||Oct 30, 1997||Aug 31, 1999||Audiotrack Corporation||Technique for embedding a code in an audio signal and for detecting the embedded code|
|US6031914||Aug 27, 1997||Feb 29, 2000||Regents Of The University Of Minnesota||Method and apparatus for embedding data, including watermarks, in human perceptible images|
|US6035177||Feb 26, 1996||Mar 7, 2000||Donald W. Moses||Simultaneous transmission of ancillary and audio signals by means of perceptual coding|
|US6061793||Aug 27, 1997||May 9, 2000||Regents Of The University Of Minnesota||Method and apparatus for embedding data, including watermarks, in human perceptible sounds|
|US6078664||Dec 20, 1996||Jun 20, 2000||Moskowitz; Scott A.||Z-transform implementation of digital watermarks|
|US6078689 *||May 18, 1998||Jun 20, 2000||Fuji Xerox Co., Ltd.||Image information encoding device|
|US6104826||Oct 9, 1997||Aug 15, 2000||Fujitsu Limited||Method of watermark-embedding/extracting identification information into/from picture data and apparatus thereof, and computer readable medium|
|US6122403||Nov 12, 1996||Sep 19, 2000||Digimarc Corporation||Computer system linked by using information in data objects|
|US6185312 *||Jan 27, 1998||Feb 6, 2001||Nippon Telegraph And Telephone Corporation||Method for embedding and reading watermark-information in digital form, and apparatus thereof|
|US6219634||Oct 14, 1998||Apr 17, 2001||Liquid Audio, Inc.||Efficient watermark method and apparatus for digital signals|
|US6226387||Aug 27, 1997||May 1, 2001||Regents Of The University Of Minnesota||Method and apparatus for scene-based video watermarking|
|US6233347||Dec 7, 1998||May 15, 2001||Massachusetts Institute Of Technology||System method, and product for information embedding using an ensemble of non-intersecting embedding generators|
|US6240121||Jul 9, 1998||May 29, 2001||Matsushita Electric Industrial Co., Ltd.||Apparatus and method for watermark data insertion and apparatus and method for watermark data detection|
|US6259738 *||Aug 8, 2000||Jul 10, 2001||Kabushiki Kaisha Toshiba||Video encoding apparatus and video decoding apparatus|
|US6272176||Jul 16, 1998||Aug 7, 2001||Nielsen Media Research, Inc.||Broadcast encoding system and method|
|US6282299||Aug 27, 1997||Aug 28, 2001||Regents Of The University Of Minnesota||Method and apparatus for video watermarking using perceptual masks|
|US6285774||Jun 8, 1998||Sep 4, 2001||Digital Video Express, L.P.||System and methodology for tracing to a source of unauthorized copying of prerecorded proprietary material, such as movies|
|US6332030||Jan 14, 1999||Dec 18, 2001||The Regents Of The University Of California||Method for embedding and extracting digital data in images and video|
|US6360000 *||Nov 9, 1998||Mar 19, 2002||David C. Collier||Method and apparatus for watermark detection for specific scales and arbitrary shifts|
|US6427012||Jun 29, 1998||Jul 30, 2002||Verance Corporation||Apparatus and method for embedding and extracting information in analog signals using replica modulation|
|US6442283||Jan 11, 1999||Aug 27, 2002||Digimarc Corporation||Multimedia data embedding|
|US6570996||Sep 2, 1998||May 27, 2003||Koninklijke Philips Electronics N.V.||Watermarking an informational signal|
|US6580809||Mar 22, 2002||Jun 17, 2003||Digimarc Corporation||Quantization-based data hiding employing calibration and locally adaptive quantization|
|US6674876||Sep 14, 2000||Jan 6, 2004||Digimarc Corporation||Watermarking in the time-frequency domain|
|US6683958||Jul 29, 2002||Jan 27, 2004||Verance Corporation||Apparatus and method for embedding and extracting information in analog signals using distributed signal features and replica modulation|
|US6751337||Aug 26, 2002||Jun 15, 2004||Digimarc Corporation||Digital watermark detecting with weighting functions|
|US7376242||Dec 19, 2003||May 20, 2008||Digimarc Corporation||Quantization-based data embedding in mapped data|
|US7454034||Jun 15, 2004||Nov 18, 2008||Digimarc Corporation||Digital watermarking of tonal and non-tonal components of media signals|
|US7769202||Aug 3, 2010||Digimarc Corporation||Quantization-based data embedding in mapped data|
|US20070237409 *||Mar 28, 2007||Oct 11, 2007||Mitsubishi Electric Corporation||Method and apparatus for compressing and decompressing images|
|GB2196167A||Title not available|
|WO1998053565A1||May 12, 1998||Nov 26, 1998||Aris Technologies, Inc.||Apparatus and method for embedding and extracting information in analog signals using distributed signal features|
|WO2000000969A1||Jun 16, 1999||Jan 6, 2000||Verance Corporation||Apparatus and method for embedding and extracting information in analog signals using replica modulation|
|WO2000022745A1||Oct 8, 1999||Apr 20, 2000||Solana Technology Development Corporation||Method and apparatus for embedding auxiliary data in primary data signal using frequency and time domain processing|
|1||Bender et al., "Techniques for Data Hiding," SPIE vol. 2420, Jan. 1995, pp. 164-173.|
|2||Boney et al, ("Digital watermarks for audio signals," Boney, L., Tewfik, A.H., Hamdy, K.N.;Multimedia Computing and Systems, 1996., Proceedings of the Third IEEE International Conference on Jun. 17-23, 1996 pp. 473-480).|
|3||Final Office Action on U.S. Appl. No. 10/869,178, mailed Apr. 6, 2007.|
|4||Goldwasser, S., et al., "Lecture Notes on Cryptography", http://www-cse.ucsd.edu/users/mihir/papers/crypto-papers.html, (Jul. 1996), 189 pages.|
|5||Haitsma et al., Audio Watermarking for Monitoring and Copy Protection, ACM Multimedia Workshop, 2000, pp. 119-122.|
|6||Hill, Simultaneous Subliminal Signalling in Conventional Sound Circuits, BBC Engineering, No. 90, May 1972, pp. 14-25.|
|7||Komatsu, "A Proposal on Digital Watermark in Document Image Communication and its Application to Realizing a Signature", Nov. 5, 1990, Electronics and Communications in Japan, Part 1, vol. 73, No. 5, pp. 22-33.|
|8||Lin, S., et al., Error Control Coding: Fundamentals and Applications, Table of Contents, (1983), 9 pages.|
|9||Matsui, "Video-Steganography: How to Secretly Embed a Signature in a Picture", Proc. Technological Strategies for Protecting Intellectual Property in the Networked Multimedia Environment, vol. 1, Issue 1, Jan. 1994, pp. 187-205.|
|10||Non-Final Office Action on U.S. Appl. No. 10/869,178, mailed Jun. 28, 2006.|
|11||Non-Final Office Action on U.S. Appl. No. 10/869,178, mailed Sep. 21, 2007.|
|12||Notice of Allowance on U.S. Appl. No. 09/228,224, mailed May 7, 2002.|
|13||Notice of Allowance on U.S. Appl. No. 10/869,178, mailed. Jul. 11, 2008.|
|14||Quatieri, T.F., et al., "Speech Transformations Based on a Sinusoidal Representation", IEEE Transactions on Acoustics, Speech, and Signal Processing, pp. 1449-1464, (Dec. 1986).|
|15||Rivest, R. L., "Cryptography", In: Hanbook of Theoretical Computer Sciences, vol. A, Ch. 13, Van Leeuwen, K., (ed.), pp. 717-755, (1990).|
|16||Rivest, R.L. "The MD4 Message Digest Algorithm", Advances in Cryptology-CRYPTO '90, pp. 303-311, (1991).|
|17||Rivest, R.L. "The MD4 Message Digest Algorithm", Advances in Cryptology—CRYPTO '90, pp. 303-311, (1991).|
|18||Schreiber et al., "A Compatible High-Definition Television System Using the Noise-Margin Mehod of Hiding Enhancement Information," SMPTE Journal, Dec. 1989, pp. 873-878.|
|19||Secure Hash Standard, Federal Information Processing Standards Publication, U.S. Department of Commerce, Technology Administration, National Institute of Standards and Technology, (Apr. 1995) 24 pages.|
|20||Swanson et al., "Data Hiding for Video-in-Video," 1997, IEEE, pp. 676-679.|
|21||Tanaka, "A Visual Retrieval System With Private Information for Image Database", Oct. 1, 1991, Proceeding Int. Conf. On DSP Applications and Technolofy, pp. 415-421.|
|22||Ten Kate, "Digital audio carrying extra information", ICASSP-90, pp. 1097-1100, Apr. 3, 1990.|
|23||Tewfik, "Data Embedding in Imagery," ARFL, Presentation, Feb. 1999, 78 pages.|
|24||Tewfik, Data Embedding Imagery and its Applications, AFRL Presentation, Jul. 1988, 47 pages.|
|25||Thompson, "Spectrum Estimation and Harmonic Analysis," Proceedings of the IEEE vol. 70, Issue 9, Sep. 1982, pp. 1055-1096.|
|26||Zhu et al., "Image Coding by Folding," Proc. Intl. Conf. on Image Processing, vol. 2, pp. 665-668, Oct. 1997.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US9099080||Jul 1, 2013||Aug 4, 2015||Muzak Llc||System for targeting location-based communications|
|US9317872||Jul 1, 2013||Apr 19, 2016||Muzak Llc||Encoding and decoding an audio watermark using key sequences comprising of more than two frequency components|
|U.S. Classification||382/100, 713/176|
|International Classification||G06K9/00, G06T1/00|
|Cooperative Classification||G06T2201/0065, G06T2201/0202, G06T1/005, G06T2201/0052, G10L19/018, G06T1/0028|
|European Classification||G06T1/00W6, G06T1/00W2|
|May 12, 2010||AS||Assignment|
Owner name: DIGIMARC CORPORATION (AN OREGON CORPORATION),OREGO
Free format text: MERGER;ASSIGNOR:DIGIMARC CORPORATION (A DELAWARE CORPORATION);REEL/FRAME:024369/0582
Effective date: 20100430
Owner name: DIGIMARC CORPORATION (AN OREGON CORPORATION), OREG
Free format text: MERGER;ASSIGNOR:DIGIMARC CORPORATION (A DELAWARE CORPORATION);REEL/FRAME:024369/0582
Effective date: 20100430
|Sep 17, 2010||AS||Assignment|
Owner name: DIGIMARC CORPORATION, OREGON
Free format text: RELEASE BY SECURED PARTY;ASSIGNOR:LEWIS, C. MCKENZIE, III;REEL/FRAME:025000/0407
Effective date: 20011217
|Oct 29, 2010||AS||Assignment|
Owner name: DMRC LLC, OREGON
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:DIGIMARC CORPORATION (A DELAWARE CORPORATION);REEL/FRAME:025217/0508
Effective date: 20080801
|Nov 2, 2010||AS||Assignment|
Owner name: DMRC CORPORATION, OREGON
Free format text: MERGER;ASSIGNOR:DMRC LLC;REEL/FRAME:025227/0808
Effective date: 20080801
Owner name: DIGIMARC CORPORATION, OREGON
Free format text: MERGER;ASSIGNOR:DMRC CORPORATION;REEL/FRAME:025227/0832
Effective date: 20080903
Owner name: COGNICITY, INC., MINNESOTA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:TEWFIK, AHMED;SWANSON, MITCHELL D.;ZHU, BIN;SIGNING DATES FROM 19990420 TO 19990428;REEL/FRAME:025227/0640
|Jul 10, 2012||CC||Certificate of correction|
|Jun 24, 2015||FPAY||Fee payment|
Year of fee payment: 4